Is the expectation at least 1 vm per tenant or start multiple nodes on the same vm to handle multiple tenants databases?
Hey @Tyler_Tallman we haven’t been focused on the multi-tenancy design aspects so far, but in principle it should be viable to make work and support eventually. In the meantime it would be interesting & useful to hear more about your requirements, e.g.:
- Are you already comfortably working with the JVM directly, or would you need a turn-key approach?
- What order of magnitude number of tenants would your application ultimately need? (10? 10,000? 10M?)
- Yes we are very comfortable working with the JVM.
- We are currently in 10s per jvm with datomic (internal testing suggests we could get to 1000s with hefty vms and using datomic client api) would like to be 100s or 1000s per jvm .
Of course tenant size varies very significantly. largest client is 50-100X the load of the smaller clients.
1 Like
Thanks for the numbers - good to understand. A follow-on question: do you need any operational ability to query across tenants with low-latency?
We do not do that very often at the moment, mostly because it is difficult. There is some value to unlock querying across tenants. We have been experimenting with streaming the data to BigQuery and doing a lot of our more complicated or cross tenant query there.
1 Like